AI Glossary
Rev’s AI speech recognition (ASR) engine recognizes a very large vocabulary, but it can still miss niche or industry-specific terms, unusual spellings, and proper nouns such as names of people, products, or places.
The AI Glossary lets you give the ASR engine a list of these words and phrases so it’s more likely to transcribe them correctly. You add terms to your account once, and Rev uses them on every English AI transcription order you place afterward. Files you’ve already transcribed aren’t changed.
The AI Glossary doesn’t apply to AI captions or human transcription. For human orders, you can add glossary terms and speaker names to each order at checkout instead.
In this article
- Getting Started
- Working with the Glossary
- Updating the Glossary
- When the Glossary Isn’t Used
- Glossary Best Practices
Getting Started
- Log in to your Rev account.
- Open the menu next to your name in the top right, expand Resources, and select Manage Glossary.
- On the Glossary for AI transcription page, enter your words or phrases in the box, one per line.
- Select Save. Your next English AI transcription orders will use the glossary.
Working with the Glossary
The following limitations currently apply to the Glossary:
- Maximum 1000 phrases
- Maximum 5 words per phrase, separated by a space
- Maximum 34 characters per word
- Only English letters (A–Z), spaces, hyphens, and apostrophes are supported. Accented letters such as é or ñ, numbers, and other symbols can’t be used.
The Glossary also includes two useful features to make data entry easier and error-free:
- You can paste a multi-line text block into the editor, and the editor will automatically format it for use, including removing empty lines and splitting comma-separated phrases.
- All words and phrases are checked as you enter them. Any entry with a problem is marked with an exclamation point (!) icon. Select the icon to see what’s wrong.
Updating the Glossary
Your glossary stays saved on your account and applies to all future English AI transcription orders. To add, change, or remove terms, go to Manage Glossary, edit the list in the box, and select Save. Changes apply only to orders you place after saving.
When the Glossary Isn’t Used
The glossary is only used on English AI transcription orders. It isn’t used in these cases:
- AI captions. To use your glossary on captions, order AI transcription instead, then download the transcript as a caption file such as SRT or VTT.
- Automatic Language Detection. If your default language in AI Transcript Settings is set to Automatic Language Detection, the glossary is turned off. Set your default language to English to turn it back on.
- Non-English files. The glossary supports English only.
- Human transcription and captions. Add terms to the order’s glossary at checkout instead.
Glossary Best Practices
To achieve the best transcription results possible with the Glossary, follow these recommendations.
Uniqueness
The Glossary is intended only for special or unusual words and phrases, such as made-up words or names with unusual spellings (for example, sparkletini, yulg, or Dave Zsykpo). Common words and well-known names, such as Google, don’t need to be added. You can’t add pronunciations, definitions, or instructions, only the spellings you want to see.
Significance
Add only important or key terms to the Glossary so that the list stays short (no more than 500 phrases). Long lists may negatively impact performance and accuracy. Remember that you can always update the Glossary to include additional words that you consider important.
Length
The Glossary limits you to 5 words per phrase because performance degrades past this number. Keep phrases as short as possible. For example, instead of the phrase "increasing awareness of outside influence", use the shorter phrase "outside influence".
Review
Review the Glossary periodically to remove terms that are no longer relevant and to add new terms based on your changing requirements. If you start to see increased errors after adding a large number of new terms to the Glossary, reduce the number of terms to the most important ones.